Skip to main content

Dr Ping Lu receives award at International Conference

Post-doctoral researcher Dr Ping Lu receives Best Poster Presentation Award at 2021 Functional Imaging and Modeling of the Heart conference

Dr Ping Lu

Post-doctoral researcher Dr Ping Lu, of the Institute of Biomedical Engineering

The 11th biennial International Conference on Functional Imaging and Modeling of the Heart, was held from June 21-25, 2021 as a fully virtual event. The conference is one of the most respected within the field of cardiovascular imaging, electrophysiology and computational modelling. As part of this conference post-doctoral researcher Dr Ping Lu (Institute of Biomedical Engineering, IBME)  discussed new developments with the most experienced cardiovascular researchers and was the recipient of the award for Best Poster Presentation, based on her research paper, ‘Multiscale Graph Convolutional Networks for Cardiac Motion Analysis’, co-authored with Professor Alison Noble, from the IBME, and Professor Daniel Rueckert and Dr Wenjai Bai from Imperial College London.

“The relationship between different areas of the heart evolves over time"

Lu explains the study, “The motion dynamics of the beating heart is a complex process that is closely related to human health. In clinical diagnosis, motion features are important biomarkers in magnetic resonance imaging (MRI) for detecting and monitoring heart disease. They are sensitive to subtle changes in myocardial function and often indicate the early onset of cardiac disease.”

Lu continues, “The relationship between different areas of the heart evolves over time. This relationship over time can be captured by an irregular graph topology instead of a regular grid.”

Before Lu’s research, attempts to restructure deep learning architectures to form  a grid-like structure to a more irregular framework had not been explored in cardiovascular image analysis.

Lu’s study proposes a convolutional neural network architecture that tracks movement spatially and over time at different scales , called a multiscale spatio-temporal graph convolutional network (MST-GCN). This MST-GCN approach is designed to learn the left ventricular (LV) motion patterns from cardiac magnetic resonance image sequences. 

Demonstration of the MST-GCN approach on the left ventricle of a heart

Each image is mapped to a network with points called nodes, which are fixed in relation to the heart, but move over time with the movement of the ventricle. This network then extracts features found at different scales and then fuses those together across scales, to model the internal relations of the nodes and form a global representation of the motion of the heart.

"For me, winning an award gives me confidence in the work I have done as well as setting a standard for my future work"

The technique allows researchers to track more than the motion of the heart, “We also show that the proposed method can estimate a number of motion-related metrics, including endocardial radii, thickness and strain which are useful for regional LV function assessment.”

“This technique can take advantage of a sparse representation of the muscular tissue of the heart using contours instead of images.”

On receiving the award, Lu says , “I am glad to receive the recognition that comes with this award. For me, winning an award gives me confidence in the work I have done as well as setting a standard for my future work. It validates my effort and encourages me to continue with my research.”



Dr Lu’s work is supported by SmartHeart. EPSRC grant EP/P001009/1. The study used data from the UK Biobank Resource under Application Number 40119.